3.1 Field of the Invention
The exemplary, illustrative, technology herein relates to thermal material processing and solid state device manufacturing. More specifically the technology herein relates to improved methods for heteroepitaxial and epitaxial growth of semiconductor materials onto a single crystal substrate or wafer using an Atomic Layer Deposition (ALD) process followed by a rapid thermal annealing step. In particular, group III/V nitride films are grown on single crystal silicon wafers in an ALD reaction chamber and rapid thermal annealing is used to restructure the deposition films to remove dislocations and reduce internal stress s.
3.2 The Related Art
GaN is an important semiconductor material usable to emit blue or violet light in Light Emitting Diodes (LED) and blue lasers. While it is highly desirable to grow single crystal GaN layers on single crystal silicon wafer substrates conventional heteroepitaxial GaN devices are constructed by growing a GaN layer onto on a sapphire substrate in part to reduce a mismatch between the crystal lattice spacing of silicon as compared to the crystal lattice spacing of GaN and further due to a mismatch between the Thermal Coefficient of Expansion (TCE) of silicon as compared GaN.
Generally single crystal sapphire substrates cost significantly more than single crystal silicon substrates in part because of lower raw material cost for silicon but also because the silicon substrate manufacture is more widely known and practiced. Unfortunately conventional wisdom continues to maintain that single crystal silicon is not as suitable as single crystal sapphire when manufacturing GaN and AN devices and more generally that for heteroepitaxial growth of any of the group III-V compounds (e.g. comprising Boron, Aluminum, Gallium, Indium and Thallium) and group II-VI compounds (e.g. comprising Cadmium and Zinc) and group III-N compounds silicon is not the best substrate choice. Instead the sapphire substrate is still in widespread use.
Thus there is a need in the art to develop a manufacturing technique suitable for heteroepitaxial growth of the group III-V compounds (e.g. comprising Boron, Aluminum, Gallium, Indium and Thallium) and or the group II-VI compounds (e.g. comprising Cadmium and Zinc) and group III-N on silicon substrates to at least take advantage of the reduced material and processing costs available afforded by use of silicon wafer based devices.
While the sapphire substrate provides good stability, low reactivity and ability to withstand the rigors of semiconductor device processing its cost, its dielectric nature and its large bandgap, preclude the possibility of fabricating devices with backside electrical contacts and have led to renewed interest in seeking alternate substrate materials with silicon being the most desirable.
3.3 Crystal Lattice Spacing Mismatch
Heteroepitaxial growth is generally more successful when the crystal lattice structure or lattice spacing of dissimilar materials is reasonably matched. This is true because the crystal lattice spacing of deposition or active layer attempts to match the lattice spacing of the substrate layer near the heteroepitaxial boundary and this action generally disorganizes the formation of the natural lattice spacing of the deposition layer to the point that the deposition layer formed at least proximate to the heteroepitaxial boundary is substantially amorphous or at best polycrystalline. In one scenario the deposition layer growth nucleates in single crystal groups forming independently but at different crystal lattice orientations with dislocations formed at the boundaries between independent single crystal formations. The dislocations further disrupt single crystal growth causing a flawed single crystal structure or a polycrystalline structure. In practice flaws in the single crystal structure potentially leading to cracking in the deposition layer especially when the deposition layer is thermally stressed during a rapid thermal cycling. The conventional solution to this problem has been to avoid heteroepitaxial devices such by growing grow silicon deposition layers on silicon substrates to avoid dislocations and encourage single crystal growth. However silicon devices have failed to provide the desired electrical properties in many applications, especially in power devices such as power switches and rectifiers. Similarly silicon devices have failed to provide the desired optical properties for optical devices. In both cases higher band gap materials such a Gallium Nitride (GaN) are desirable as deposition or active layers and preferably the GaN are most economically formed on silicon substrates.
Single crystal sapphire is a single crystal form of corundum (Al2O3) also known as alpha aluminum, alumina. Sapphire's crystal structure is a hexagonal system, rhomboidal class 3 m which makes single crystal sapphire more compatible as a substrate for the growth of single crystal or nearly single crystal III-V compounds (e.g. comprising Boron, Aluminum, Gallium, Indium and Thallium) than silicon. Additionally group II-VI compounds (e.g. comprising Cadmium and Zinc) and III-N have similar crystal structures to the II-V compounds such that single crystal sapphire substrates are more compatible for heteroepitaxial growth of these compounds than single crystal silicon substrates. In particular silicon and GaN have a crystal lattice mismatch of 16.9% while sapphire and GaN have a crystal lattice mismatch of 13.62% providing a slight advantage to sapphire.
It is highly desirable for single crystal formation (long term order) in active deposition layers in order to provide homogeneous electrical and or optical properties. Specifically this means forming a substantially uniform crystal lattice orientation throughout the layer volume and the more uniform the crystal lattice orientation of the active deposition layer the better the electrical and optical properties of the eventual semiconductor device is likely to be. In the case of laser and laser diode devices comprising a gallium nitride active layer, better crystal orientation leads to increased luminous intensity at the device output and a narrower spectral bandwidth of the output radiation wherein substantially all the spectral output is at the primary spectral response of the device.
3.4 Thermal Cycling and Thermal Coefficient of Expansion Mismatch
It is widely accepted that single crystal heteroepitaxial layers can only be formed at an epitaxial growth temperature Tg which for GaN is reported to be at least 550° C. See e.g. Trivedi et al. Low-temperature GaN growth on silicon substrates by single gas-source epitaxy and photo-excitation; Appl. Phys. Lett. 87, 072107 (2005). It is known to manufacture semiconductor lasers and LED's by forming GaN layers onto a substantially single crystal sapphire substrate using a conventional Metal Organic Chemical Vapor Deposition (MOCVD) process. In particular it is generally accepted that the highest quality conventional GaN devices are fabricated when the deposition temperature is in the range of 900 to 1100° C. However even in the highest quality devices some crystal lattice defects caused by crystal lattice spacing mismatch are inevitable at the heteroepitaxial boundary.
Moreover even with the effort taken to match the TCE of the sapphire substrate with the GaN the high temperature range of 900 to 1100° C. required in a conventional MOCVD reactor necessitates strict thermal stress management in order to avoid excessive wafer bow and cracking resulting from thermal cycling with non-matched TCE materials. Typically wafer bow is limited to less than less than 100 μm in order for the wafer to be further processed on conventional wafer handling and processing tools for high volume manufacturing. While wafer bow has been addressed by forming “stress compensation layers” in MOCVD films, these stress compensation layers degrade the device layer performance and add cost.
3.5 Use of an Aluminum Nitride (AlN) Transition Layer
Recently Pan et al. (Growth of GaN film on Si(111) Substrate using a AlN sandwich structure as buffer Joun. Of Crystal Growth 318 (2011) 464-467) report attempts to grow device quality GaN onto a Si (111) substrate. In this example a sandwich structure formed by an AlN nucleation layer formed on a Si(111) substrate followed by mixed AlN/GaN transition layers and finally an active GaN layer were all formed by metal-organic chemical vapor deposition (MOCVD) at 1060° C. in attempt to reduce undesirable cracking due to the large crystal lattice spacing mismatch (16.9%) between the GaN and the silicon and the large coefficient of thermal expansion (CTE) mismatch between GaN (αa 5.59×10−6 K−1) and Si(αa 3.77×10−6 K−1). While it was hoped that the sandwich structure would alleviate surface cracks the results were disappointing. While Pan et al. report that the GaN epitaxial layers grew uniformly on Si substrates the active layer suffered from randomly distributed cracks, which they report are mostly caused by the CTE mismatch.
3.6 Diffusion Across the Heteroepitaxial Boundary
One drawback of the growing films on substrates at high temperature (above about 800° C.) is related to diffusion that can occur at the layer boundaries. In particular substrate nitridation can occur at the high deposition temperatures. Additionally the extreme thermal gradients and thermal cycling range can lead to some cracking in the films and the substrates. In one particular example gallium nitride layers are grown with a high concentration of nitrogen vacancies at conventional MOCVD temperatures of 900 to 1100° C. The nitrogen vacancies lead to a high background carrier concentration in the device thus degrading electrical and electro-optical properties.
While attempts have been made to grow the GaN films at lower temperatures while still using a MOCVD process on sapphire; films grown at 500° C. have a 1000 times weaker photoluminescence than films grown at 800° C. (reference). Another drawback of high reaction temperature MOCVD processing is that indium and some other group III-V and group II-VI compounds have a thermal stability than prevents their use above 800° C. thus limiting the extent to which high reaction temperature MOCVD processing can be used to deposit some group III-V and group II-VI compounds on any substrate. Thus there is a need in the art to develop a lower temperature deposition technique.
Experimenters have attempted to deposit GaN onto single crystal silicon substrates with some success. However, the films that have formed generally include mixed crystal lattice orientations (i.e. they are not single crystals) even when the silicon substrate is a single crystal. This is due in part to the mismatch in crystal lattice structure between silicon and gallium nitride. Unfortunately the resulting LED and laser devices are not competitive with conventional devices manufactured with single crystal lattice orientation on single crystal sapphire.
3.7 Growing Heteroepitaxial Layers Using Atomic Layer Deposition
Atomic Layer Deposition (ALD) systems are available that can deposit material layers at lower deposition temperatures e.g. 80-550° C. and techniques and precursors suitable for depositing both GaN and AlN onto a silicon substrate by an ALD process are known and discloses in the below listed references.
Kim and co-workers in Atomic layer deposition of GaN using GaCl3 and NH3 (J. Vac. Sci. Technol. A 27, 4, July/August 2009), have grown GaN on Si(100) substrates by ALD in a temperature window of 500-700° C. The halogenated precursors GaCl, and GaCl3, were used with NH3 as the co-reactant on a silicon substrate. The exposure time of the GaCl3 precursor was varied over the range of 2-7 seconds. The results indicate mixed crystallographic orientations of GaN, e.g. a mixture of (0002) and (1011) oriented GaN and high Cl-content in the films which would be detrimental for device applications.
The X-ray Diffraction (XRD) results shown in FIG. 4A plots (410) and (420) are from films grown via ALD at 550-650 C for a 25 sec exposure time and a 7 sec exposure time respectively. Note the mixed crystal orientation, and the weak crystallinity in plot (420) formed with a lower exposure time as compared to plot (410) formed with a longer exposure time. By comparison, the XRD results in FIG. 4B are from high temperature MOCVD grown GaN (grown at 900-1100° C.). Note the clear monocrystalline nature of the MOCVD film grown at 900-1100° C. which exhibits a predominantly GaN(0002) crystal structure.
Ozgit and coworkers (Proceedings of the E-MRS Fall Meeting, Symposium H: Warsaw, Poland, Sep. 19_23, 2011), have grown GaN on Si(100) substrates by plasma-enhanced atomic layer deposition using trimethyl gallium (TMG) and trimethyl gallium (TEG) and ammonia (NH3) precursors in a temperature window of 250-350° C. for TMG and 150-350° C. for TEG. No crystallinity measurements are provided. They report the films exhibited linear growth behavior with an oxygen content of 19.5 to 22.5%.
AlN, while traditionally being grown via MOCVD or sputtering to produce device quality films, has also been attempted via ALD, though to a very small extent.
Results from Alevli and co-workers—in The Influence of Growth Temperature on the Properties of AN Films Grown by ALD (Proceedings of the E-MRS Fall Meeting, Symposium H: Warsaw, Poland, Sep. 19-23, 2011), using TMA and NH3 as the precursor and co-reactant indicates high temperature growth is not feasible due to thermal decomposition of the precursors—this in turn leading to rougher films at higher temperatures. Crystallinity is claimed at low temperatures (100-200 C), but no results are presented.
Other results on growth of AN via ALD has been shown by Liu and co-workers (ECS Transactions, 2011) to produce mixed crystalline results even at process temperatures up to 400 C—however, with progressively weakening (101) orientation, and a strengthening of the desired (002) orientation, as the temperature increases, (See present FIG. 5).
There is a need in the art to deposit GaN onto silicon substrates. Growing GaN devices onto Si potentially decreases the cost of these devices considerably. However, the lattice structure of Si is sufficiently different than the lattice structure of GaN, so that, dislocations are abundant and single crystal layers of GaN are not formed. There have been many attempts to circumvent this problem by depositing onto Si wafers that are cut so that the lattice constant of the exposed surface is more similar to the lattice constant of GaN. Other attempts have been through depositing an AlN buffer layer onto Si and then growing the GaN layer onto the AN layer. The AlN acts as a buffer layer and relaxes the stresses in the GaN film caused by the different lattice constants between GaN and Si. However, a significant problem is that all of the deposition processes are at elevated temperatures. The thermal expansion coefficient of Aluminum Nitride (AlN) and GaN are different than Si, leading to growth defects during the film deposition process due to the Thermal Coefficient of Expansion (TCE) mismatch.
It would be desirable to be able to grow the GaN directly on a Si substrate at or near room temperature, or to grow an AlN buffer layer onto Si, followed by GaN active layer on top of the AlN at low temperatures, e.g. at or below 350° C.